High resolution mass spectrometry for quantitative analysis and untargeted screening of algal

25 Pages·2015·1.1 MB·English

1

Journal of Chromatography A Achimer
O ctober 2015, Volume 1416, Pages 10-21
http://archimer.ifremer.fr
http://dx.doi.org/10.1016/j.chroma.2015.08.064
http://archimer.ifremer.fr/doc/00277/38848/
© 2015 Elsevier B.V. All rights reserved.

High resolution mass spectrometry for quantitative analysis
and untargeted screening of algal toxins in mussels and
passive samplers
Zendong Zita 1, 3, *, Mccarron Pearse 2, Herrenknecht Christine 3, Sibat Manoella 1, Amzil Zouher 1,
Cole Richard B. 4, Hess Philipp 1

1 Ifremer, Laboratoire Phycotoxines, Rue de l’Ile d’Yeu, 44311 Nantes, France
2 National Research Council of Canada, Biotoxin Metrology, Measurement Science and Standards,
1411 Oxford St, Halifax, Nova Scotia B3H 3Z 1, Canada
3 LUNAM, Université de Nantes, MMS EA2160, Faculté de Pharmacie, 9 rue Bias, 44035 Nantes,
France
4 Institut Parisien de Chimie Moléculaire, UMR 8232, Université Pierre et Marie Curie (Paris VI), 4 Place
Jussieu, 75252 Paris, France
* Corresponding author : Zita Zendong, email address : [email protected]

Abstract :

Measurement of marine algal toxins has traditionally focussed on shellfish monitoring while, over the
last decade, passive sampling has been introduced as a complementary tool for exploratory studies.
Since 2011, liquid chromatography–tandem mass spectrometry (LC–MS/MS) has been adopted as the
EU reference method (No. 15/2011) for detection and quantitation of lipophilic toxins. Traditional LC–MS
approaches have been based on low-resolution mass spectrometry (LRMS), however, advances in
instrument platforms have led to a heightened interest in the use of high-resolution mass spectrometry
(HRMS) for toxin detection. This work describes the use of HRMS in combination with passive sampling
as a progressive approach to marine algal toxin surveys. Experiments focused on comparison of LRMS
and HRMS for determination of a broad range of toxins in shellfish and passive samplers.
Matrix effects are an important issue to address in LC–MS; therefore, this phenomenon was evaluated
for mussels (Mytilus galloprovincialis) and passive samplers using LRMS (triple quadrupole) and HRMS
(quadrupole time-of-flight and Orbitrap) instruments. Matrix-matched calibration solutions containing
okadaic acid and dinophysistoxins, pectenotoxin, azaspiracids, yessotoxins, domoic acid, pinnatoxins,
gymnodimine A and 13-desmethyl spirolide C were prepared. Similar matrix effects were observed on
all instruments types. Most notably, there was ion enhancement for pectenotoxins, okadaic
acid/dinophysistoxins on one hand, and ion suppression for yessotoxins on the other. Interestingly, the
ion selected for quantitation of PTX2 also influenced the magnitude of matrix effects, with the sodium
adduct typically exhibiting less susceptibility to matrix effects than the ammonium adduct. As expected,
mussel as a biological matrix, quantitatively produced significantly more matrix effects than passive
sampler extracts, irrespective of toxin. Sample dilution was demonstrated as an effective measure to
reduce matrix effects for all compounds, and was found to be particularly useful for the non-targeted
Please note that this is an author-produced PDF of an article accepted for publication following peer review. The definitive
publisher-authenticated version is available on the publisher Web site.
2
approach.
Limits of detection and method accuracy were comparable between the systems tested, demonstrating
the applicability of HRMS as an effective tool for screening and quantitative analysis. HRMS offers the
advantage of untargeted analysis, meaning that datasets can be retrospectively analyzed. HRMS (full
scan) chromatograms of passive samplers yielded significantly less complex data sets than mussels,
and were thus more easily screened for unknowns. Consequently, we recommend the use of HRMS in
combination with passive sampling for studies investigating emerging or hitherto uncharacterized toxins.

Highlights
► Quantitative HRMS-method developed for targeted screening of biotoxins. ► Advantage of HRMS
over LRMS with regards to untargeted screening of unknowns. ► Similar magnitude and direction of
matrix effects in HRMS compared to LRMS. ► Less matrix effects with passive sampler matrix
compared to mussel matrix.

Keywords : Monitoring, Marine toxins, Passive sampling, SPATT, Matrix effects

Please note that this is an author-produced PDF of an article accepted for publication following peer review. The definitive
publisher-authenticated version is available on the publisher Web site.
3

53 1. INTRODUCTION
54 A number of micro-algae produce marine toxins that can be accumulated in filter-feeding shellfish
55 species such as mussels and oysters, and thus lead to human intoxication through consumption [1]. For
56 several decades, the complexity of the toxins produced by these algae has impeded method
57 development due to the lack of reference calibrants and materials. Therefore, generic mouse bioassays
58 were often used, despite commonly accepted drawbacks [2]. Liquid chromatography coupled to
59 tandem mass spectrometry (LC-MS/MS) has become a versatile tool for the analysis otf food and
60 environmental contaminants, including toxins. LC-MS/MS is now the reference mpethod for the
61 detection and quantitation of toxins produced by harmful algae [3]. To achieve this goal, different
62 studies have developed and validated quantitative methods for the analysis of phyicotoxins, typically
r
63 using low resolution mass spectrometry (LRMS) [4-9]. This technique is now being increasingly used
c
64 for monitoring [10, 11] and for characterization of reference materials [12, 13]. Additionally, methods
65 using high resolution mass spectrometry (HRMS) have recently been desveloped and quantitatively
66 validated for some marine toxins [14-16].
67 However, an important issue to address when developing or validatingu a quantitative analytical method
68 using LC-MS via electrospray (ESI) and atmospheric pressure ionization (API) sources is the possible
n
69 occurrence of matrix effects [17, 18]. Matrix effects are considered to be an alteration in analyte
70 response due to the presence of co-eluting compounds, either due to mass interference (isobaric
a
71 compounds) or alteration of the desorption/ionization efficiency due to co-elution. These co-eluting
72 compounds may increase (ion enhancement) or reduceM (ion suppression) the desorption/ionization of
73 the targeted analyte [19, 20]. Matrix effects may arise from different co-eluting components:
74 endogenous compounds already present as sample constituents and still present after extraction or

75 sample pre-treatment, or from reagents added to the mobile phase to improve chromatographic
d
76 separation and peak shape [21], as well as from interfering materials used during extraction procedures
77 or even from variable elution flow-rates [22]. Matrix effects can be easily detected when comparing
e
78 the response obtained from standard solutions to those from spiked matrix extracts. In the presence of
79 matrix effects, both identification atnd determination of analytes can be affected [22]. Therefore, the
80 evaluation of matrix effects in MpS detection and solutions to overcome them should be examined in
81 the early stages of development of new methods. Several approaches have been used to alleviate
82 matrix effects in the quantitaetive analysis of lipophilic marine toxins. These approaches include SPE
83 cleanup and column flushing [23, 24], matrix-matched calibration and standard addition [24-26],
c
84 reduction of the injection volume [11], use of an internal standard and use of a different ionization
85 source such as APCcI [19].
86 For applications that require analyses of complex biological samples, the use of HRMS can offer at
A
87 least two major advantages: (i) the ability to overcome mass interferences stemming from overlapping
88 signals of isobaric species (at low resolution such interferences lead to overestimation of the quantity
89 of the analyte present) and (ii) non-targeted screening (where mass spectrometry is used to survey the
90 contents of a complex mixture). In the field of toxins a good example of HRMS dealing with
91 interfering isobaric compounds is the case of anatoxin-a, which may be hampered by the presence of
92 phenylalanine [27]. HRMS has also been the prime technique for non-targeted screening of complex
93 samples for unknowns, employing Orbitrap and Time-of-Flight mass spectrometers [9, 28, 29].
94 While monitoring of biotoxins has traditionally been carried out in mussels, passive samplers, also
95 referred to as Solid Phase Adsorption Toxin Tracking (SPATT) have been more recently introduced to
96 detect toxins in the marine environment [30]. Subsequently, many studies have successfully
97 implemented passive sampling, using mainly the HP20 resin, to detect lipophilic toxins in different
98 aquatic environments [31-35]. This technique has not yet proven to be useful as a monitoring tool for
99 early warning of harmful algal blooms [36]. However, passive samplers have the advantage that unlike
Page 3 of 25
4

100 in mussels, the adsorbed toxins do not undergo biotransformation. Mussels have traditionally been
101 used in many monitoring programs since they can be classified as a sentinel species due to the
102 relatively unselective feeding of mussels compared to other bivalve mollusks, e.g. oysters.
103 In this study, we evaluate and compare matrix effects caused by mussel matrix and passive sampler
104 components in the analysis of different phycotoxins, using both low and high resolution mass
105 spectrometers. As a complement to the overall non-targeted approach employing HRMS, a range of
106 toxins was investigated quantitatively: from relatively hydrophilic toxins such as domoic acid (DA)
107 and yessotoxins (YTX and homo-YTX), over toxins of intermediate lipophilicity such as tpinnatoxins
108 E, F and G (PnTX-E, -F, -G), gymnodimine A (GYM-A), 13-desmethylspirolide-C (13-dpesmeSPX-C),
109 to the more lipophilic ones including azaspiracids 1 to 3 (AZA1, -2, -3), okadaic acid (OA)
i
110 dinophysistoxins 1 and 2 (DTX1, -2), pectenotoxin 2 (PTX2) and brevetoxin-1 arnd 2 (BTX1, -2). A
111 chromatographic separation method was developed and optimized to obtain good separation of the
c
112 toxins of interest. Matrix matched calibration curves, prepared using mussel and passive sampler
113 extracts, were injected on different analytical systems with low resolution s(triple quadrupole) and high
114 resolution (orbitrap and quadrupole time-of-flight) mass spectrometers. The impact of the ion selected
u
115 for quantitation, sample dilution and use of low or high resolution detectors on matrix effects were
116 assessed. Finally, the study evaluated the benefits of passive samnpler matrix as a complementary tool
117 to traditionally used shellfish matrix (mussels) with the help of HRMS for an untargeted, exploratory
118 approach. a
119
120 2. EXPERIMENTAL M
121

122 2.1. Chemicals and reagents d
123 Certified calibration solutions were frome the National Research Council of Canada (NRCC, Halifax,
124 NS, Canada). These included calibration solution CRMs: domoic acid (DA), azaspiracids 1, 2 and 3
t
125 (AZA1-3), pectenotoxin 2 (PTX2), okadaic acid (OA) dinophysistoxins 1 and 2 (DTX1 and -2),
p
126 yessotoxin (YTX), homo-yessotoxin (homo-YTX), 13-desmethyl spirolide C (13-desmeSPX-C),
127 pinnatoxin G (PnTX-G) and gymnodimine A (GYM-A); and mussel tissue CRMs: CRM-ASP-Mus-d,
e
128 CRM-DSP-Mus-c and CRM-AZA-Mus. A multitoxin tissue material CRM-FDMT-1 undergoing
129 certification, well-characcterized in-house calibration solutions for PnTX-E and F, brevetoxins 1 and 2
130 (BTX1 and -2), 20-methyl spirolide G (20-me-SPX-G) and pectenotoxin-2-seco acid (PTX2sa), as
c
131 well as a mussel extract (Bruckless, Donegal, Ireland – 2005) containing different azaspiracids were
132 also provided bAy NRC.
133 Alexandrium ostenfeldii (A. ostenfeldii) extract containing 13,19-didesmethyl spirolide C (13,19-
134 didesme-SPX-C) and Ostreopsis ovata (O. ovata) extract containing ovatoxin a (OvTX-a) were
135 obtained from Ifremer as previously described [33, 37]. Those extracts were mixed with some of the
136 abovementioned certified and in-house reference toxin calibration solutions as well as the mussel
137 extract from Bruckless to obtain a composite multi-toxin sample, used for optimization of
138 chromatographic separation.
139 HPLC-grade methanol, acetonitrile and formic acid (98%) were obtained from Sigma Aldrich
140 (Steinheim, Germany) and Caledon (Georgetown, ON, Canada). Ammonium formate was from Fluka
141 (St. Louis, MI, USA). Milli-Q water was produced in-house at 18MΩ/cm quality, using a Milli-Q
142 integral 3 system (Millipore). For analyses with HRMS instruments, acetonitrile and water of LCMS-
143 grade were obtained from Fisher Scientific (Illkirch, France).
144
Page 4 of 25
5

145
146
146 2.2. Instrumentation and analytical methods
147
148 2.2.1. LC-MS/MS systems t
p
149 System A: Triple quadrupole (QqQ):
i
150 An Agilent HPLC 1100 series system (1.58 min dwell time) was connected to an API4000™ mass
r
151 spectrometer (AB Sciex) equipped with a TurboIonSpray™ ionization source. For quantitation, the
c
152 mass spectrometer was operated in MRM mode, scanning two transitions for each toxin. Q1 and Q3
153 resolutions of the instrument were set at Unit (arbitrary terms). Data were ascquired in scheduled MRM
154 and the target scan time was 1 s in both positive and negative modes. MRM detection windows were
u
155 set at 45 s in both polarities. Data acquisition was carried out with Analyst 1.6 Software (AB Sciex).
156 Optimized parameters are shown in Table 1.
n
157
158 Table 1: Optimized transitions selected for scheduled MRMa method.
Q3 Q3
Toxin DP [V] Q1 M CE [eV] CE [eV]
quantifier qualifier
DA 61 312.1 266.1 23 161.1 35
OvTX-a 65 1315.7 327.1 45 1298 25
GYM-A 90 508.4d 490.2 30 392.3 50
13,19-didesMe-C 120 678.5 430.5 45 164.5 65
e
13-desmeSPX-C 90 692.5 164.1 70 444.2 60
20-me-SPX-G 85 706.6 164.1 70 346.3 50
t
PnTX-G 125 694.5 164.1 80 458.3 60
p
PnTX-E 125 784.5 164.1 80 766.5 60
PnTX-F e125 766.5 164.1 80 748.5 60
AZA1 60 842.5 672.4 65 362.3 75
c
AZA2 60 856.5 672.4 65 362.3 75
AZA3 60 828.5 658.4 65 362.3 75
c
AZA6 110 842.5 658.4 65 362.3 75
A
AZA33 110 716.5 698.5 40 362.4 70
AZA34 116 816.5 798.4 41 672.5 69
BTX1 70 884.6 221.1 35 403.4 30
BTX2 90 912.5 895.5 19 877.5 29
PTX2 80 876.5 823.5 35 213.1 55
PTX2sa 85 894.5 823.5 35 213.1 60
OA, DTX2 -80 803.5 255.1 -65 113.1 -85
DTX1 -70 817.5 255.1 -70 113.1 -90
YTX -70 1141.6 1061.6 -55 855.5 -70
homo-YTX -70 1155.6 1075.6 -55 869.5 -70
159
160 System B: Quadrupole Time of Flight (Q-ToF):
Page 5 of 25
6

161 A UHPLC system (1290 Infinity II, Agilent Technologies, Waldbronn, Germany) with a 0.3 min dwell
162 time was coupled to a 6550 iFunnel QToF (Agilent Technologies, Santa Clara, CA, USA) equipped
163 with a dual ESI source. This instrument was operated with a dual electrospray ion source with Agilent
164 Jet Stream Technology™ in positive (ESI+) and negative (ESI-) ionization modes. Mass spectra were
165 acquired over the scan range m/z 100 - 1200 with an acquisition rate of 0.5 s. The parameters of the Jet
166 Stream Technologies™ source in ESI+ were: gas temperature 205 °C, drying gas flow 16 L/min,
167 nebulizer pressure 50 psig, sheath gas temperature 355 °C, sheath 12 L/min, capillary voltage 2 kV,
168 fragmentor voltage, 200 V. In ESI- the parameters were as follows: gas temperature 290 °Ct, drying gas
169 flow 12 L/min, nebulizer pressure 50 psig, sheath gas temperature 355 °C, sheath 12 L/pmin, capillary
170 voltage -3.5 kV, fragmentor voltage, 180 V. The instrument was mass calibrated in positive and
i
171 negative ionization modes before each analysis, using the Agilent tuning mrix. Additionally, all
172 experiments were carried out using reference mass correction using purine (m/z 121.05087 [M+H]+;
c
173 m/z 119.03632 [M-H]-) and HP-921 = hexakis(1H,1H,3H-tetrafluoropropoxy) phosphazine (m/z
174 922.00979 [M+H]+; m/z 966.00072 [M+HCOO]-). The reference ions weres infused constantly with an
175 isocratic pump to a separate ESI sprayer in the dual spray source. MassHunter Acquisition B05.01
u
176 software was used to control the instrument and data were processed with MassHunter B07.00.
n
177 System C: Orbitrap:
178 Analyses were carried out on an Accela™ High Speed LaC (dwell time 0.75 min) coupled to an
179 Exactive™ mass spectrometer (ThermoFisher Scientific, Whaltham, MA, USA), equipped with an
M
180 Orbitrap mass analyzer and a heated electrospray ionization probe (HESI-II). The instrument was
181 operated and mass calibrated in positive and negative ionization modes as described previously [14].
182 “Balanced” automatic gain control (AGC) was used for all analyses, with a maximum injection time
183 set to 50 ms across a scan range of m/z 100 d- 1500. Data acquisition was carried out with Xcalibur
184 software (ThermoFisher Scientific).
185 Optimal ion source and interface conditioens consisted of a spray voltage of 3 kV (positive mode) or -
186 2.7 kV (negative mode), sheath gas flow rate of 50 (ESI+) and 25 (ESI-), auxiliary gas flow rate of 10,
t
187 capillary temperature of 360°C and heater temperature of 250°C. Acquisitions were made in full scan
p
188 with high collision dissociation (HCD) using an energy of 60 eV. Full scan and HCD data were
189 acquired at high (50000) and medium (10000) resolutions respectively. Alternative full scan and HCD
e
190 data were obtained at a scan rate of 2Hz, resulting in an overall cycle time of ca. 1 s.
c
191
c
192 2.2.2. Liquid chromatography conditions
A
193 Three different Kinetex stationary phases (C18, XB-C18 and Biphenyl from Phenomenex) of identical
194 geometry and particle size have been initially evaluated (see supplementary material Table S1). The
195 column finally selected was a Phenomenex Kinetex XB-C18 (100 x 2.1 mm; 2.6 µm).
196 The binary mobile phase consisted of (A) 100% water and (B) 95% acetonitrile. All phases contained
197 2 mM ammonium formate and 50 mM formic acid. The final gradient selected after optimization of
198 chromatographic separation used a flow-rate of 400 µL/min, and acetonitrile in the organic
199 component. The elution gradient rose from 5% to 50% of B in 3.6 min, then 100% B was reached by
200 8.5 min. After 1.5 min of hold time at 100% B, 5% B was reached within 10 s, followed by 5 min re-
201 equilibration of the column at 5% B. The total chromatographic run time was 15 min. For all
202 experiments the column temperature was maintained at 40 °C and injection volumes were 3µL. This
203 gradient was used to compare the chromatographic separation between columns in the triple
204 quadrupole system and also to assess matrix effects in all three mass spectrometry systems listed
205 above.
Page 6 of 25
7

206
207 2.3. Sample preparation
208
209 2.3.1. Mussel, passive samplers and CRMs extraction protocol
210 Mussel (Mytilus galloprovincialis) and blank HP-20 passive samplers (300 mg) used to prepare
211 matrix-matched calibration solutions had been deployed over the same 1-week period at Vtillefranche-
212 sur-mer bay (France). Mussels were prepared according to the EURLMB SOP [38] by pextracting 2 g
213 of homogenized mussels with 2 × 9 mL of 100% MeOH. After centrifugation, the supernatants were
i
214 combined into a volumetric flask and the volume adjusted to 20 mL using MeOrH. Passive samplers
215 were prepared and extracted as described [33]. SPATTs were prepared fromc HP20 resin (300 mg)
216 contained between sheets of mesh that were hold together by embroidery rings. After retrieval, each
217 SPATT was rinsed with deionized water, the resin transferred to an empsty SPE cartridge and eluted
218 with 15 mL of MeOH. Since the procedure for the preparation of matrix-matched standard required
u
219 diluting the matrix extract to 3/4 of the original volume, initial blank extracts were concentrated to 4/3
220 of the original volume under a gentle stream of nitrogen, to yieldn appropriate matrix concentration in
221 the final matrix-matched solutions. A protocol adapted from McCarron et al [13] was used to extract
222 CRMs samples. CRM material (2 g) was serially extracteda four times with 5.5 mL of MeOH. The
223 supernatants were collected and brought to 25 mL into a volumetric flask.
M
224
225 2.3.2. Matrix-matched calibration solutions for the evaluation of matrix effects

226 Due to potential stability problems of AZAs, PnTX-E and PTX2 in acidic conditions [39-41] (and the
d
227 acid present in the certified calibrant to enhance storage capacity of 13-desmeSPX-C), three initial
228 toxin mixtures were prepared in methanoel: (i) Mix-1 containing PTX2, AZA1 to 3, OA, DTX1 and 2,
229 PnTX-E, YTX, homo-YTX and DA; (ii) Mix-2 containing 13-desmeSPX-C, GYM-A, PnTX-F,
230 PnTX-G and DA and (iii) BTX1,2t-mix with BTX1 and BTX2. These stock solutions were then
231 serially diluted in MeOH using pa Hamilton Microlab diluter-dispenser (Hamilton Company, Reno,
232 NV). The samples from the serial dilution series were spiked into previously prepared and
e
233 concentrated blank mussel and SPATT extracts (from section 2.3.1): firstly, aliquots of extract (225
234 µL) were dispensed into HPLC vials, then 75 µL of each dilution level solution was added. This
c
235 operating procedure resulted in a consistent matrix concentration at each concentration level. Matrix-
236 free samples were cprepared similarly, using pure methanol instead of mussel or passive sampler
237 extracts.
A
238 The calibration curves thus covered a range from approximately 0.07 ng mL-1 to 50 ng mL-1 for AZAs
239 and okadaic acid groups, 0.04 ng mL-1 to 26 ng mL-1 for cyclic imines, 0.3 ng mL-1 to 220 ng mL-1 for
240 YTXs, 1.5 ng mL-1 to 1070 ng mL-1 for DA, 11 ng mL-1 to 740 ng mL-1 for BTX1 and 25 ng mL-1 to
241 1620 ng mL-1 for BTX2. Based on triplicate injections of seven points methanol and matrix-matched
242 calibration curves, mean slopes, intercept and correlation coefficients (R2) were calculated by
243 application of least squares adjustment without weighting.
244 Matrix effects were evaluated on the QqQ, the Q-ToF and on the Orbitrap using the Phenomenex
245 Kinetex XB-C18 (100 x 2.1 mm; 2.6 µm) column with the optimized gradient.
246
247
248 2.4. Method performance characteristics
Page 7 of 25
8

249 To assess method performances and matrix effects, each concentration for each calibration curve was
250 injected in triplicate, alternating between standards in methanol, standards in SPATT matrix and
251 standards in mussel matrix. After the injection of each matrix-matched calibration curve, a check
252 standard sample containing the monitored toxins was injected in-between two blank injections. This
253 procedure led to injection sequences of approximately 100 injections. Drift correction, if necessary,
254 was applied before any further data processing: evaluation of linearity, accuracy, matrix effects, etc.
255 (supplementary material S1).
256 Mass-to-charge ratio on high resolution instruments and the corresponding standard devitations were
257 calculated from triplicate injections of methanol, SPATT or mussel calibration soplutions. Mass
258 extraction was made with a mass accuracy window of ± 5 ppm. To avoid positive and negative errors
i
259 cancelling each other out when calculating errors (ppm) [42], absolute values ofr the individual mass
260 errors were used.
c
261 As there is not always sufficient noise to calculate signal-to-noise ratios in HRMS, detection limits
262 (LoD) were determined with the ordinary least-squares regression data smethod [43, 44] using the
263 lowest 3 points from the calibration curves (in MeOH, SPATT and mussel extracts). The LoD was
u
264 calculated as 3 times the standard deviation of the y-intercepts, over the slope of the calibration curve
265 [43, 44]. n
266 To evaluate the accuracy of the method on all three systems (QqQ, Q-ToF and Orbitrap), certified
267 reference materials containing targeted toxins at known coancentrations were analyzed: CRM-ASP-
268 mus-d for DA; CRM-DSP-mus-c for OA, DTX1 and -2; CRM-AZA-mus-d for AZA-1, -2 and -3 and
269 CRM-FDMT-1 for 13-desme-SPX-C and PTX2. M
270
271 2.5. Data treatment

d
272 Statistical evaluations were carried out using SigmaPlot 12.5. Significance tests used to compare
273 matrix effects between different conditions were a t-test, a Wilcoxon signed rank test and an ANOVA
e
274 on ranks according to Friedman using repeated measures. Differences were considered significant at p
275 < 0.05. t
276 The Agilent Molecular Feature Epxtractor (MFE) algorithm was used to obtain the Total Compound
277 Chromatogram of samples [45]. This algorithm is designed for use with full scan data and treats all of
278 the mass spectral data as a tehree-dimensional array of retention time, m/z and abundance values. At
279 this stage, any point corresponding to persistent or slowly-changing background is removed from that
c
280 array of values. Subsequently, the algorithm searches for ion traces (= Features) that have common
281 elution profile, i.e. cion traces that elute at very nearly the same retention times. Those ion traces are
282 then grouped into entities called Compounds regrouping all ion traces that are related, i.e. those that
A
283 correspond to mass peaks in the same isotope cluster, or can be explained as being different adducts or
284 charge states of the same entity. The results for each detected Compound are a mass spectrum
285 containing the ions with the same elution time and explainable relationships, and an extracted
286 compound chromatogram (ECC) computed using all of these related ion traces in the compound
287 spectrum (and only those traces). Finally, all Compounds eluting at very nearly the same retention time
288 are grouped into compound groups to facilitate data reduction. Indeed, the algorithm does not allow
289 for regrouping of true fragments different from adducts or isotopic clusters, and thus two or more of
290 the entities called Compounds from a same group may actually be derived from in-source
291 fragmentation of a single molecule.
292 Non-targeted analysis of field samples often show more complex blanks as all ionisable compounds
293 from the solvents and additives used in extraction, sample preparation and mobile phases, as well as
294 ghost-peaks from previous injections, may appear in the mass analyzer. Thus, some samples were
295 blank-subtracted post-acquisition for evaluation of data complexity. For this blank-subtraction, a
Page 8 of 25
9

296 database was constituted with all peaks that appeared in solvent blanks and HP20 (passive sampler =
297 SPATT matrix) extraction blanks. When using the MFE™ algorithm described above, an exclusion list
298 may be added to exclude these compounds present in the blank from those extracted into total
299 compound chromatograms (TCCs). Whenever blank subtraction was applied this is specifically
300 mentioned in the result and discussion section.
301
302 3. RESULTS AND DISCUSSION t
p
303
i
304 3.1. Method Selection and Performance r
c
305 Initial chromatographic method development focused on achieving good separations within the OA
306 group to avoid quantitation errors (different toxicity of OA and DTX2). Dsuring method development
307 Kinetex C18, Kinetex XB-C18 and Kinetex Biphenyl columns (100 x 2.1 mm; 2.6 µm) were
308 compared (Supplementary material Table S1). Better resolutions betwueen neighboring peaks (Rs>2)
309 were obtained on Kinetex C18 and XB-C18 compared to the Kinetex Biphenyl (supplementary
n
310 material Table S2). Of note AZA3 and PTX2 were resolved on the Kinetex XB-C18 column (Rs=6.9)
311 but not on the Kinetex C18, probably due to the slightlya higher polarity of the Kinetex XB-C18
312 column, as well as its different steric interactions [46].
313 A methanol-based mobile phase was also tested wiMth the same gradient on the three columns.
314 Methanol has a selectivity different to that of acetonitrile, and use of the same gradient led to more co-
315 elution between toxins, regardless of the column stationary phase, in particular the type of bonding

316 (supplementary material Table S2). Therefore, the mobile phase with methanol was discarded for
d
317 further experiments. However, it is noteworthy that better sensitivity was obtained for BTXs when
318 using the methanol mobile phase, compaered to the acetonitrile mobile phase (supplementary material
319 Figure S1).
320 The column finally chosen was the tKinetex XB-C18, with resolutions of Rs=6.9 between PTX2 and
321 AZA3, Rs=4.5 between OA and DpTX2 and Rs=3.2 between YTX and OA. PnTX-F and PnTX-G were
322 barely baseline resolved (Rs=2), but significant co-elution remained for BTX2 and AZA2 (Rs=1.1) in
e
323 positive ionization, and for YTX and homo-YTX in negative ion mode (supplementary material Table
324 S2 and Figure S2). We aimed to develop a relatively short method for a multiclass screening of
c
325 phycotoxins. Figure 1 shows the LC separation of 29 different algal toxins using the optimized
326 gradient. LRMS ancd HRMS approaches for multi-toxin determination were examined further using
327 these conditions for a reduced set of toxins as certified calibration solutions were not available for all
A
328 toxins.
329
Page 9 of 25
10

ESI+
t
p
i
r
c
s
u
n
a
M
330

d
ESI- e
t
p
e
c
c
A
331
332 Figure 1: HPLC chromatogram in ESI+ and ESI-, of the composite multi-toxin sample (section 2.1)
333 acquired on the Orbitrap using the Phenomenex Kinetex XB-C18 (100 x 2.1 mm; 2.6µm) with the
334 optimized gradient (acetonitrile).
Page 10 of 25